Analysis of Risk Factors for Perioperative Transfusion in Hip Arthroplasty and Modeling of a Nomogram

To investigate the risk factors of perioperative transfusion in hip arthroplasty and to construct a nomogram model to predict the risk of transfusion.

A total of 120 patients who underwent hip arthroplasty in our hospital from January 2022 to October 2023 were retrospectively selected and classified as training set. According to the same selection criteria, another 50 patients who underwent hip arthroplasty from December 2023 to December 2024 were selected as validation set. All patients' general information, surgical indicators, laboratory indicators, etc., were collected. Logistic regression analysis was performed to identify risk factors for perioperative transfusion. A risk warning scoring model-based nomogram was constructed using Rstudio 4.2.1. The predictive accuracy of the model was evaluated through internal and external validation.

Among the 120 patients in the training set, 34 of them (28.33%) who received perioperative transfusion were categorized into transfusion group, in which 19 patients (15.83%) received intraoperative transfusion whereas 15 (12.50%) received postoperative transfusion. Patients who did not received perioperative transfusion were categorized into non-transfusion group (n = 86). Compared with subjects in the non-transfusion group, those in the transfusion group were elder, experienced longer surgery, suffered from more severe intraoperative bleeding, and had lower hemoglobin level (p < 0.05). Multivariate regression analysis showed that age, operation time, and intraoperative hemorrhage were the independent factors, and that hemoglobin level was an independent protective factor for perioperative transfusion in hip arthroplasty. Receiver operating characteristic (ROC) curve analysis showed that the area under the curve (AUC) for the combined prediction of age, operation time, intraoperative bleeding, and hemoglobin in the context of perioperative transfusion during hip arthroplasty was 0.940, which was higher than that of individual prediction. The difference between the general information of the training set and validation set was not statistically significant (p > 0.05). The calibration curves of the nomogram for predicting perioperative transfusion in hip arthroplasty converged to the ideal curve (training set: p = 0.651, validation set: p = 0.181). Decision curve analysis (DCA) curves depict a higher positive net benefit for the nomogram in the probability threshold range of 5% to 97%. The AUC of the ROC curve for predicting perioperative transfusion in hip arthroplasty in the training and validation sets were 0.940 (0.916-0.991) and 0.925 (0.879-0.941), respectively.

Age, operation time and intraoperative bleeding are independent risk factors, while hemoglobin level is an independent protective factor for perioperative transfusion in hip arthroplasty. The nomogram established based on the above factors has good predictive performance, which is helpful in assessing the risk of perioperative transfusion, formulating personalized transfusion strategies, reducing unnecessary transfusions and related complications.